next up previous contents
Next: Quantile-quantile plot Up: Graphical representation Previous: Boxplot (box-and-whiskers plot)   Contents

Histogram

The range of values is divided up into a finite set of class intervals (bins). The number of objects in each bin is then counted and divided by the sample size to obtain the frequency of occurrence and then these are plotted as vertical bars of varying height. It is also possible to divide the frequencies by the bin width to obtain frequency densities that can then be compared to probability densities from theoretical distributions. For example, a suitably scaled normal probability density function has been superimposed on the frequency histogram in Figure 2.3.

Figure: Histogram of sample heights showing frequency in each bin with suitably scaled normal density curve superimposed.
\begin{figure}\centerline{
\epsfysize=8cm
\epsffile{dbsfigs/histogramn.eps}
}\end{figure}

The histogram quickly reveals the location, spread, and shape of the distribution. The shape of the distribution can be unimodal (one hump), multimodal (many humps), skewed (fatter tail to left or right), or more-peaked and fatter tails (leptokurtic), or less-peaked and thinner tails (platykurtic) than a normal (Gaussian) distribution.



David Stephenson 2005-09-30